Static-first, evidence-aware
Imports, strings, function indices, decompilation, and triage reports are cached and reused so the analysis gets richer instead of noisier.
GhostTrace combines a cyberpunk web workspace, a Ghidraaas backend, an Ollama-driven analyst, and a reproducible Windows lab with SSH and x64dbg bridge tooling. It is designed for fast static triage, guided deep dives, and dynamic evidence workflows without losing the thread between tools.
Most reverse engineering setups degenerate into tabs, folders, shell history and screenshots. GhostTrace tries to collapse that into one loop: upload a sample, let the assistant triage it, preserve evidence, and carry that context into debugging and sandbox work.
Imports, strings, function indices, decompilation, and triage reports are cached and reused so the analysis gets richer instead of noisier.
The Windows lab ships with noVNC, RDP, SSH, OEM provisioning, x64dbg autostart bridge and host-side helpers for command execution and file transfer.
The web app is built around jobs, triage, debugger state and guided next steps, with a design language that looks deliberate instead of generic.
The shape is inspired by good offensive and defensive RE workflows: start broad, shrink uncertainty fast, keep the expensive steps targeted, and always preserve what you learned for the next pass.
The Flask web UI hashes the sample, forwards it to Ghidraaas, and tracks the resulting analysis job in a persistent sidebar.
GhostTrace generates structured reports from imports, strings, functions and optional evidence so the assistant can recommend a sensible next move immediately.
The x64dbg bridge exposes debugger state, queued requests and findings back into the same workspace, instead of sending you off to a disconnected side tool.
The Windows VM is provisioned reproducibly, reachable over SSH, and fed through host-side copy helpers so you can script it instead of babysitting it.
No. GhostTrace is the workbench that orchestrates them. The point is to keep context, triage and debugger evidence flowing through one operator-facing surface.
No. The default path is Ollama running on the host. The Windows sandbox does not run its own model stack by default.
No. The main stack works without it. The sandbox profile is optional and only matters when you want the Windows-side tooling and debugger bridge.
Because upstream packaging is inconsistent on Windows for a few tools. The lab now completes successfully even when `r2ai` or `PE-bear` are unavailable, instead of failing the whole bootstrap.
Technically, yes. GhostTrace is intended for legitimate reverse engineering, malware analysis, DFIR, research, and defensive engineering. If someone points it at unauthorized targets or uses it for illicit work, that responsibility sits with the operator. Keep it sharp, not stupid.